🔍 StackSearch — AI-Powered Developer Documentation Assistant

Screen.Recording.2026-01-05.at.11.51.31.mov

StackSearch is an AI-powered search assistant that lets developers query framework documentation using natural language. Ask anything about AWS, Docker, FastAPI, React, MongoDB, Node.js, Redis, PostgreSQL, Tailwind CSS, TypeScript, and more — StackSearch retrieves the most relevant docs using vector search (FAISS + Gemini embeddings) and replies only from the official documentation.

If the answer isn’t in the docs, StackSearch will strictly reply:

Not in docs.

This project was built as a real-world demonstration of: ✅ AI-powered RAG (Retrieval-Augmented Generation)
✅ FAISS similarity search
✅ Gemini Embeddings + LLM
✅ Multi-framework doc ingestion & indexing
✅ React frontend with markdown & code highlighting
✅ MongoDB chat history persistence
✅ Google Authentication

---

✨ Features

🧠 AI-Powered Documentation Search

• Ask natural-language questions about any supported tech stack
• Answers strictly grounded in documentation
• Prevents hallucinations using “Not in docs” fallback

📚 Multi-Framework Support

Supporting (and expandable): AWS • Docker • Express.js • FastAPI • MongoDB • Next.js • NumPy • Pandas PostgreSQL • Prisma • React • Redis • Tailwind • TypeScript Node.js • GraphQL • Kubernetes • Go • Python • Django

🧩 Vector Search using FAISS

• Docs split into semantic chunks
• Embedded with gemini-embedding-001
• Stored in a monolithic FAISS index
• Filtered per-framework using normalized metadata

🔐 Google Login (Firebase Auth)

• Secure login gateway
• Only authenticated users can chat

💾 Persistent Chat History (MongoDB)

• Messages stored per-framework
• Conversations restored on reload

💬 Clean Chat UI

• Built with React + Tailwind
• Markdown rendering
• Syntax-highlighted code blocks
• One-click copy button
• Beautiful dark theme

---

🏗️ Tech Stack

🔹 Frontend

• React
• TypeScript
• Tailwind CSS
• React Markdown
• Syntax Highlighter
• Redux Toolkit
• Firebase Auth
• React Router

🔹 Backend

• Node.js
• Express
• TypeScript
• MongoDB + Mongoose
• LangChain
• Gemini LLM + Embeddings
• FAISS Vector Index

---

🚀 Getting Started

1️⃣ Clone the Repo

git clone https://github.com/yourusername/stacksearch.git
cd stacksearch

2️⃣ Install Dependencies (Backend)

cd server
npm install

3️⃣ Set up environment variables

cp .env.example .env

4️⃣ Ingest docs

npm run ingest

5️⃣ Run the app

npm run dev

6️⃣ Install Dependencies (Frontend)

cd client
npm install

7️⃣ Set up environment variables

cp .env.example .env

8️⃣ Run the app

npm run dev

🔑 Authentication Flow

• User logs in with Google via Firebase
• Redux stores user session
• Logged-in users can access chat
• Logout clears state + storage

🗄️ Chat History Storage

• Each framework maintains its own conversation thread.

Schema:

 chatId: string       // framework name
 framework: string
 history: [
  { role: "user" | "assistant", content: string, timestamp: Date }
]

RAG Search Flow

• User asks question
• Query embedded via Gemini
• FAISS retrieves top-k relevant doc chunks
• Gemini LLM answers ONLY using those docs
• If missing → reply "Not in docs."
• Save messages to MongoDB